Harshvardhan P

Data Scientist – Graduate Engineer Trainee

L&T Finance

Jul 2025 - Present

Identified a high-potential segment within the 90+ DPD delinquent base using machine learning classification; achieved a 5.6% recovery rate vs. a portfolio average of 0.3% – an 18× improvement. Collaborated across analytics, policy, MRM and business teams to drive model approval and production deployment. Monitored and retrained propensity models predicting Top-up Loan uptake: an on-us model leveraging internal customer features and an off-us model using credit bureau variables for MFI borrowers. Enabled targeted campaigns capturing ∼70% conversion within the top 30% of the offer population across a ∼Rs. 1500 Cr monthly disbursal portfolio. Delivered data-driven insights and ad-hoc analyses supporting critical policy decisions for Top-up Loan and Micro Business Loan products; findings directly influenced underwriting and targeting strategy. Refactored and migrated legacy R/SQL pipelines for the pre-approved pool generation process to Python, reducing server load and processing time. Ensured all policy changes were accurately implemented, version-controlled, and auditable.